telldir()/seekdir() portability fixes

POSIX says that an offset obtained from telldir() can only be used
on the same DIR *. Linux is abls to reuse the offset accross
closedir()/opendir() for a given directory, but this is not portable
and such a behavior should be fixed.

An incomplete fix for the posix xlator was merged in
http://review.gluster.com/8926
This change set completes it.

- Perform the same fix index xlator.
- Use appropriate casts and variable types so that 32 bit signed
  offsets obtained by telldir() do not get clobbered when copied into
  64 bit signed types.
- modify glfs-heal.c and afr-self-heald.c so that they do not use
  anonymous fd, since this will cause closedir()/opendir() between
  each syncop_readdir(). On failure we fallback to anonymous fs
  only for Linux so that we can cope with updated client vs not
  updated brick.
- Avoid sending an EINVAL when the client request for the EOF offset.
  Here we fix an error in previous fix for posix xlator: since we
  fill each directory entry with the offset of the next entry, we
  must consider as EOF the offset of the last entry, and not the
  value of telldir() after we read it.
- Add checks in regression tests that we do not hit cases where
  offsets fed to seekdir() are wrong. Introduce log_newer() shell
  function to check for messages produced by the current script.

This fix gather changes from http://review.gluster.org/9047
and http://review.gluster.org/8936 making them obsolete.

BUG: 1129939
Change-Id: I59fb7f06a872c4f98987105792d648141c258c6a
Signed-off-by: Emmanuel Dreyfus <manu@netbsd.org>
Reviewed-on: http://review.gluster.org/9071
Tested-by: Gluster Build System <jenkins@build.gluster.com>
Reviewed-by: Raghavendra Bhat <raghavendra@redhat.com>
Tested-by: Raghavendra Bhat <raghavendra@redhat.com>

9 files changed, 137 insertions, 31 deletions

diff --git a/heal/src/glfs-heal.c b/heal/src/glfs-heal.c
index 1056ea467c4..9416316458d 100644
--- a/heal/src/glfs-heal.c
+++ b/heal/src/glfs-heal.c
@@ -284,7 +284,7 @@ _get_afr_ancestor (xlator_t *xl)
 
 static int
 glfsh_process_entries (xlator_t *xl, fd_t *fd, gf_dirent_t *entries,
-                       off_t *offset, uint64_t *num_entries)
+                       uint64_t *offset, uint64_t *num_entries)
 {
         gf_dirent_t      *entry = NULL;
         gf_dirent_t      *tmp = NULL;
@@ -359,7 +359,7 @@ out:
 static int
 glfsh_crawl_directory (xlator_t   *readdir_xl, fd_t *fd, loc_t *loc)
 {
-        off_t           offset   = 0;
+        uint64_t        offset = 0;
         gf_dirent_t     entries;
         int             ret = 0;
         gf_boolean_t    free_entries = _gf_false;
@@ -472,7 +472,27 @@ glfsh_print_pending_heals (xlator_t *xl, loc_t *rootloc)
                 goto out;
         }
 
-        fd = fd_anonymous (dirloc.inode);
+        fd = fd_create (dirloc.inode, GF_CLIENT_PID_GLFS_HEAL);
+        if (!fd) {
+                printf ("fd_create failed: %s", strerror(errno));
+                goto out;
+        }
+        ret = syncop_opendir (xl, &dirloc, fd);
+        if (ret) {
+                fd_unref(fd);
+#ifdef GF_LINUX_HOST_OS /* See comment in afr_shd_index_opendir() */
+                fd = fd_anonymous (dirloc.inode);
+                if (!fd) {
+                        printf ("fd_anonymous failed: %s",
+                                strerror(errno));
+                        goto out;
+                }
+#else
+                printf ("opendir failed: %s", strerror(errno));
+                goto out;
+#endif
+        }
+
         ret = glfsh_crawl_directory (xl, fd, &dirloc);
         if (fd)
                 fd_unref (fd);
diff --git a/libglusterfs/src/common-utils.h b/libglusterfs/src/common-utils.h
index 2493ba10816..23a932b71aa 100644
--- a/libglusterfs/src/common-utils.h
+++ b/libglusterfs/src/common-utils.h
@@ -115,6 +115,7 @@ enum _gf_client_pid
         GF_CLIENT_PID_NO_ROOT_SQUASH    = -4,
         GF_CLIENT_PID_QUOTA_MOUNT       = -5,
         GF_CLIENT_PID_AFR_SELF_HEALD    = -6,
+        GF_CLIENT_PID_GLFS_HEAL         = -7,
 };
 
 typedef enum _gf_boolean gf_boolean_t;
diff --git a/tests/basic/afr/self-heald.t b/tests/basic/afr/self-heald.t
index 37e10b254c1..6937cf227d7 100644
--- a/tests/basic/afr/self-heald.t
+++ b/tests/basic/afr/self-heald.t
@@ -4,6 +4,7 @@
 . $(dirname $0)/../../volume.rc
 
 cleanup;
+START_TIMESTAMP=`date +%s`
 
 function disconnected_brick_count {
         local vol=$1
@@ -162,4 +163,7 @@ TEST $CLI volume create $V0 $H0:$B0/${V0}{6}
 TEST $CLI volume start $V0
 TEST ! $CLI volume heal $V0 info
 
+# Check for non Linux systems that we did not mess with directory offsets
+TEST ! log_newer $START_TIMESTAMP "offset reused from another DIR"
+
 cleanup
diff --git a/tests/basic/pump.t b/tests/basic/pump.t
index 6e81ddde6cd..e9e54a7a9ea 100644
--- a/tests/basic/pump.t
+++ b/tests/basic/pump.t
@@ -4,6 +4,7 @@
 . $(dirname $0)/../volume.rc
 
 cleanup;
+START_TIMESTAMP=`date +%s`
 
 TEST glusterd
 TEST pidof glusterd
@@ -41,4 +42,7 @@ done
 
 EXPECT "" echo $files
 
+# Check for non Linux systems that we did not mess with directory offsets
+TEST ! log_newer $START_TIMESTAMP "offset reused from another DIR"
+
 cleanup
diff --git a/tests/include.rc b/tests/include.rc
index 75db2d10d0e..7c31eb65402 100644
--- a/tests/include.rc
+++ b/tests/include.rc
@@ -783,6 +783,24 @@ function MKFS_LOOP ()
   esac
 }
 
+# usage: log_newer timestamp "string"
+# search in glusterfs logs for "string" logged after timestamp seconds
+# since the Epoch (usually obtained by date +%s)
+log_newer()
+{
+        ts=$1
+        msg=$2
+        logdir=`$CLI --print-logdir`
+
+        IFS="["
+        for date in `grep -hr "$msg" $logdir | awk -F '[\]]' '{print $1}'` ; do
+                if [ `date -d "$date" +%s` -gt $ts ] ; then
+                        return 0;
+                fi
+        done 2>/dev/null
+        return 1
+}
+
 function MOUNT_LOOP ()
 {
   if [ $# != 2 ] ; then
diff --git a/xlators/cluster/afr/src/afr-self-heald.c b/xlators/cluster/afr/src/afr-self-heald.c
index a341015303c..29f34107481 100644
--- a/xlators/cluster/afr/src/afr-self-heald.c
+++ b/xlators/cluster/afr/src/afr-self-heald.c
@@ -274,6 +274,8 @@ afr_shd_index_opendir (xlator_t *this, int child)
 #ifdef GF_LINUX_HOST_OS
 		fd_unref (fd);
 		fd = fd_anonymous (inode);
+		if (!fd)
+		        goto out;
 #else /* GF_LINUX_HOST_OS */
 		gf_log(this->name, GF_LOG_ERROR,
 		       "opendir of %s for %s failed: %s",
@@ -427,7 +429,7 @@ afr_shd_index_sweep (struct subvol_healer *healer)
 	fd_t *fd = NULL;
 	xlator_t *subvol = NULL;
 	afr_private_t *priv = NULL;
-	off_t offset = 0;
+	uint64_t offset = 0;
 	gf_dirent_t entries;
 	gf_dirent_t *entry = NULL;
 	uuid_t gfid;
@@ -501,11 +503,12 @@ afr_shd_index_sweep (struct subvol_healer *healer)
 int
 afr_shd_full_sweep (struct subvol_healer *healer, inode_t *inode)
 {
+	loc_t loc = {0, };
 	fd_t *fd = NULL;
 	xlator_t *this = NULL;
 	xlator_t *subvol = NULL;
 	afr_private_t *priv = NULL;
-	off_t offset = 0;
+	uint64_t offset = 0;
 	gf_dirent_t entries;
 	gf_dirent_t *entry = NULL;
 	int ret = 0;
@@ -514,9 +517,38 @@ afr_shd_full_sweep (struct subvol_healer *healer, inode_t *inode)
 	priv = this->private;
 	subvol = priv->children[healer->subvol];
 
-	fd = fd_anonymous (inode);
-	if (!fd)
-		return -errno;
+	uuid_copy (loc.gfid, inode->gfid);
+	loc.inode = inode_ref(inode);
+
+	fd = fd_create (inode, GF_CLIENT_PID_AFR_SELF_HEALD);
+	if (!fd) {
+		gf_log(this->name, GF_LOG_ERROR,
+		       "fd_create of %s failed: %s",
+		       uuid_utoa (loc.gfid), strerror(errno));
+		ret = -errno;
+		goto out;
+	}
+
+	ret = syncop_opendir (subvol, &loc, fd);
+	if (ret) {
+#ifdef GF_LINUX_HOST_OS /* See comment in afr_shd_index_opendir() */
+		fd_unref(fd);
+		fd = fd_anonymous (inode);
+		if (!fd) {
+			gf_log(this->name, GF_LOG_ERROR,
+			       "fd_anonymous of %s failed: %s",
+			       uuid_utoa (loc.gfid), strerror(errno));
+			ret = -errno;
+			goto out;
+		}
+#else /* GF_LINUX_HOST_OS */
+		gf_log(this->name, GF_LOG_ERROR,
+		       "opendir of %s failed: %s",
+		       uuid_utoa (loc.gfid), strerror(errno));
+		ret = -errno;
+		goto out;
+#endif /* GF_LINUX_HOST_OS */
+	}
 
 	INIT_LIST_HEAD (&entries.list);
 
@@ -558,6 +590,8 @@ afr_shd_full_sweep (struct subvol_healer *healer, inode_t *inode)
 			break;
 	}
 
+out:
+	loc_wipe (&loc);
 	if (fd)
 		fd_unref (fd);
 	return ret;
@@ -947,7 +981,7 @@ afr_shd_gather_index_entries (xlator_t *this, int child, dict_t *output)
 	fd_t *fd = NULL;
 	xlator_t *subvol = NULL;
 	afr_private_t *priv = NULL;
-	off_t offset = 0;
+	uint64_t offset = 0;
 	gf_dirent_t entries;
 	gf_dirent_t *entry = NULL;
 	uuid_t gfid;
diff --git a/xlators/features/index/src/index.c b/xlators/features/index/src/index.c
index 5892f741629..f7408c1308f 100644
--- a/xlators/features/index/src/index.c
+++ b/xlators/features/index/src/index.c
@@ -268,10 +268,11 @@ check_delete_stale_index_file (xlator_t *this, char *filename)
 }
 
 static int
-index_fill_readdir (fd_t *fd, DIR *dir, off_t off,
+index_fill_readdir (fd_t *fd, index_fd_ctx_t *fctx, DIR *dir, off_t off,
                     size_t size, gf_dirent_t *entries)
 {
         off_t     in_case = -1;
+        off_t     last_off = 0;
         size_t    filled = 0;
         int       count = 0;
         char      entrybuf[sizeof(struct dirent) + 256 + 8];
@@ -286,11 +287,11 @@ index_fill_readdir (fd_t *fd, DIR *dir, off_t off,
         } else {
                 seekdir (dir, off);
 #ifndef GF_LINUX_HOST_OS
-                if (telldir(dir) != off) {
+                if ((u_long)telldir(dir) != off && off != fctx->dir_eof) {
                         gf_log (THIS->name, GF_LOG_ERROR,
-                                "seekdir(%ld) failed on dir=%p: "
+                                "seekdir(0x%llx) failed on dir=%p: "
 				"Invalid argument (offset reused from "
-				"another DIR * structure?)", (long)off, dir);
+				"another DIR * structure?)", off, dir);
                         errno = EINVAL;
                         count = -1;
                         goto out;
@@ -299,7 +300,7 @@ index_fill_readdir (fd_t *fd, DIR *dir, off_t off,
         }
 
         while (filled <= size) {
-                in_case = telldir (dir);
+                in_case = (u_long)telldir (dir);
 
                 if (in_case == -1) {
                         gf_log (THIS->name, GF_LOG_ERROR,
@@ -335,12 +336,13 @@ index_fill_readdir (fd_t *fd, DIR *dir, off_t off,
                 if (this_size + filled > size) {
                         seekdir (dir, in_case);
 #ifndef GF_LINUX_HOST_OS
-                        if (telldir(dir) != in_case) {
+                        if ((u_long)telldir(dir) != in_case &&
+                            in_case != fctx->dir_eof) {
 				gf_log (THIS->name, GF_LOG_ERROR,
-					"seekdir(%ld) failed on dir=%p: "
+					"seekdir(0x%llx) failed on dir=%p: "
 					"Invalid argument (offset reused from "
 					"another DIR * structure?)",
-					(long)in_case, dir);
+					in_case, dir);
 				errno = EINVAL;
 				count = -1;
 				goto out;
@@ -357,7 +359,14 @@ index_fill_readdir (fd_t *fd, DIR *dir, off_t off,
                                 entry->d_name, strerror (errno));
                         goto out;
                 }
-                this_entry->d_off = telldir (dir);
+                /*
+                 * we store the offset of next entry here, which is
+                 * probably not intended, but code using syncop_readdir()
+                 * (glfs-heal.c, afr-self-heald.c, pump.c) rely on it
+                 * for directory read resumption.
+                 */
+                last_off = (u_long)telldir(dir);
+                this_entry->d_off = last_off;
                 this_entry->d_ino = entry->d_ino;
 
                 list_add_tail (&this_entry->list, &entries->list);
@@ -366,9 +375,12 @@ index_fill_readdir (fd_t *fd, DIR *dir, off_t off,
                 count ++;
         }
 
-        if ((!readdir (dir) && (errno == 0)))
+        if ((!readdir (dir) && (errno == 0))) {
                 /* Indicate EOF */
                 errno = ENOENT;
+                /* Remember EOF offset for later detection */
+                fctx->dir_eof = last_off;
+        }
 out:
         return count;
 }
@@ -581,6 +593,7 @@ __index_fd_ctx_get (fd_t *fd, xlator_t *this, index_fd_ctx_t **ctx)
                 fctx = NULL;
                 goto out;
         }
+        fctx->dir_eof = -1;
 
         ret = __fd_ctx_set (fd, this, (uint64_t)(long)fctx);
         if (ret) {
@@ -950,7 +963,7 @@ index_readdir_wrapper (call_frame_t *frame, xlator_t *this,
                 goto done;
         }
 
-        count = index_fill_readdir (fd, dir, off, size, &entries);
+        count = index_fill_readdir (fd, fctx, dir, off, size, &entries);
 
         /* pick ENOENT to indicate EOF */
         op_errno = errno;
@@ -1271,8 +1284,11 @@ index_releasedir (xlator_t *this, fd_t *fd)
                 goto out;
 
         fctx = (index_fd_ctx_t*) (long) ctx;
-        if (fctx->dir)
-                closedir (fctx->dir);
+        if (fctx->dir) {
+                ret = closedir (fctx->dir);
+                if (ret)
+                        gf_log (this->name, GF_LOG_ERROR, "closedir error: %s", strerror (errno));
+        }
 
         GF_FREE (fctx);
 out:
diff --git a/xlators/features/index/src/index.h b/xlators/features/index/src/index.h
index 661dcdbc417..206d280e519 100644
--- a/xlators/features/index/src/index.h
+++ b/xlators/features/index/src/index.h
@@ -34,6 +34,7 @@ typedef struct index_inode_ctx {
 
 typedef struct index_fd_ctx {
         DIR *dir;
+        off_t dir_eof;
 } index_fd_ctx_t;
 
 typedef struct index_priv {
diff --git a/xlators/storage/posix/src/posix.c b/xlators/storage/posix/src/posix.c
index a3aa57a0153..638e0f59202 100644
--- a/xlators/storage/posix/src/posix.c
+++ b/xlators/storage/posix/src/posix.c
@@ -4857,6 +4857,7 @@ posix_fill_readdir (fd_t *fd, DIR *dir, off_t off, size_t size,
                     gf_dirent_t *entries, xlator_t *this, int32_t skip_dirs)
 {
         off_t     in_case = -1;
+        off_t     last_off = 0;
         size_t    filled = 0;
         int             count = 0;
         char      entrybuf[sizeof(struct dirent) + 256 + 8];
@@ -4892,11 +4893,11 @@ posix_fill_readdir (fd_t *fd, DIR *dir, off_t off, size_t size,
         } else {
                 seekdir (dir, off);
 #ifndef GF_LINUX_HOST_OS
-                if (telldir(dir) != (long)off && off != pfd->dir_eof) {
+                if ((u_long)telldir(dir) != off && off != pfd->dir_eof) {
                         gf_log (THIS->name, GF_LOG_ERROR,
-                                "seekdir(%ld) failed on dir=%p: "
+                                "seekdir(0x%llx) failed on dir=%p: "
                                 "Invalid argument (offset reused from "
-                                "another DIR * structure?)", (long)off, dir);
+                                "another DIR * structure?)", off, dir);
                         errno = EINVAL;
                         count = -1;
                         goto out;
@@ -4905,7 +4906,7 @@ posix_fill_readdir (fd_t *fd, DIR *dir, off_t off, size_t size,
         }
 
         while (filled <= size) {
-                in_case = telldir (dir);
+                in_case = (u_long)telldir (dir);
 
                 if (in_case == -1) {
                         gf_log (THIS->name, GF_LOG_ERROR,
@@ -4965,13 +4966,13 @@ posix_fill_readdir (fd_t *fd, DIR *dir, off_t off, size_t size,
                 if (this_size + filled > size) {
                         seekdir (dir, in_case);
 #ifndef GF_LINUX_HOST_OS
-                        if (telldir(dir) != (long)in_case &&
+                        if ((u_long)telldir(dir) != in_case &&
                             in_case != pfd->dir_eof) {
                                 gf_log (THIS->name, GF_LOG_ERROR,
-                                        "seekdir(%ld) failed on dir=%p: "
+                                        "seekdir(0x%llx) failed on dir=%p: "
                                         "Invalid argument (offset reused from "
                                         "another DIR * structure?)",
-                                        (long)in_case, dir);
+                                        in_case, dir);
                                 errno = EINVAL;
                                 count = -1;
                                 goto out;
@@ -4988,7 +4989,14 @@ posix_fill_readdir (fd_t *fd, DIR *dir, off_t off, size_t size,
                                 entry->d_name, strerror (errno));
                         goto out;
                 }
-                this_entry->d_off = telldir (dir);
+                /*
+                 * we store the offset of next entry here, which is
+                 * probably not intended, but code using syncop_readdir()
+                 * (glfs-heal.c, afr-self-heald.c, pump.c) rely on it
+                 * for directory read resumption.
+                 */
+                last_off = (u_long)telldir(dir);
+                this_entry->d_off = last_off;
                 this_entry->d_ino = entry->d_ino;
                 this_entry->d_type = entry->d_type;
 
@@ -5002,7 +5010,7 @@ posix_fill_readdir (fd_t *fd, DIR *dir, off_t off, size_t size,
                 /* Indicate EOF */
                 errno = ENOENT;
                 /* Remember EOF offset for later detection */
-                pfd->dir_eof = telldir (dir);
+                pfd->dir_eof = (u_long)last_off;
         }
 out:
         return count;